
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import warnings
warnings.filterwarnings("ignore")

from sklearn.feature_selection import SelectKBest
from sklearn.feature_selection import chi2
from sklearn.preprocessing import StandardScaler

from sklearn.model_selection import train_test_split
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import KFold

from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from sklearn.naive_bayes import GaussianNB
from xgboost import XGBClassifier
from sklearn.ensemble import RandomForestClassifier

from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
from sklearn.metrics import f1_score


url_train_label = 'https://raw.githubusercontent.com/ReynaraEzra/Richter-Earthquake-Prediction/main/train_labels.csv'
url_train_value = 'https://raw.githubusercontent.com/ReynaraEzra/Richter-Earthquake-Prediction/main/train_values.csv'
url_test_value = 'https://raw.githubusercontent.com/ReynaraEzra/Richter-Earthquake-Prediction/main/test_values.csv'

train_label = pd.read_csv(url_train_label)
train_value = pd.read_csv(url_train_value)
test_value = pd.read_csv(url_test_value)


# train_label = pd.read_csv("train_labels.csv")
# train_value = pd.read_csv("train_values.csv")
# test_value = pd.read_csv("test_values.csv")


print('Train Values : {}'.format(train_value.shape))
print('Train Labels : {}'.format(train_label.shape))
print('Test Values : {}'.format(test_value.shape))

Train Values : (260601, 39)
Train Labels : (260601, 2)
Test Values : (86868, 39)


print('Number of Training Data : {}'.format(train_value.shape[0]))
print('Number of Testing Data : {}'.format(test_value.shape[0]))
print('Total Data : {}'.format(train_value.shape[0]+test_value.shape[0]))

Number of Training Data : 260601
Number of Testing Data : 86868
Total Data : 347469


train_label.head()


train_value.head()


test_value.head()


train = pd.merge(train_value, train_label)


train.head()


train.tail()


train.sample(5)


train.isnull().sum()

building_id                               0
geo_level_1_id                            0
geo_level_2_id                            0
geo_level_3_id                            0
count_floors_pre_eq                       0
age                                       0
area_percentage                           0
height_percentage                         0
land_surface_condition                    0
foundation_type                           0
roof_type                                 0
ground_floor_type                         0
other_floor_type                          0
position                                  0
plan_configuration                        0
has_superstructure_adobe_mud              0
has_superstructure_mud_mortar_stone       0
has_superstructure_stone_flag             0
has_superstructure_cement_mortar_stone    0
has_superstructure_mud_mortar_brick       0
has_superstructure_cement_mortar_brick    0
has_superstructure_timber                 0
has_superstructure_bamboo                 0
has_superstructure_rc_non_engineered      0
has_superstructure_rc_engineered          0
has_superstructure_other                  0
legal_ownership_status                    0
count_families                            0
has_secondary_use                         0
has_secondary_use_agriculture             0
has_secondary_use_hotel                   0
has_secondary_use_rental                  0
has_secondary_use_institution             0
has_secondary_use_school                  0
has_secondary_use_industry                0
has_secondary_use_health_post             0
has_secondary_use_gov_office              0
has_secondary_use_use_police              0
has_secondary_use_other                   0
damage_grade                              0
dtype: int64


print('Ukuran data sebelum drop duplicate :', train.shape)
train = train.drop_duplicates(subset='building_id', keep='last')
print('Ukuran data setelah drop duplicate :', train.shape)

Ukuran data sebelum drop duplicate : (260601, 40)
Ukuran data setelah drop duplicate : (260601, 40)


train.columns

Index(['building_id', 'geo_level_1_id', 'geo_level_2_id', 'geo_level_3_id',
       'count_floors_pre_eq', 'age', 'area_percentage', 'height_percentage',
       'land_surface_condition', 'foundation_type', 'roof_type',
       'ground_floor_type', 'other_floor_type', 'position',
       'plan_configuration', 'has_superstructure_adobe_mud',
       'has_superstructure_mud_mortar_stone', 'has_superstructure_stone_flag',
       'has_superstructure_cement_mortar_stone',
       'has_superstructure_mud_mortar_brick',
       'has_superstructure_cement_mortar_brick', 'has_superstructure_timber',
       'has_superstructure_bamboo', 'has_superstructure_rc_non_engineered',
       'has_superstructure_rc_engineered', 'has_superstructure_other',
       'legal_ownership_status', 'count_families', 'has_secondary_use',
       'has_secondary_use_agriculture', 'has_secondary_use_hotel',
       'has_secondary_use_rental', 'has_secondary_use_institution',
       'has_secondary_use_school', 'has_secondary_use_industry',
       'has_secondary_use_health_post', 'has_secondary_use_gov_office',
       'has_secondary_use_use_police', 'has_secondary_use_other',
       'damage_grade'],
      dtype='object')


train.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 260601 entries, 0 to 260600
Data columns (total 40 columns):
 #   Column                                  Non-Null Count   Dtype 
---  ------                                  --------------   ----- 
 0   building_id                             260601 non-null  int64 
 1   geo_level_1_id                          260601 non-null  int64 
 2   geo_level_2_id                          260601 non-null  int64 
 3   geo_level_3_id                          260601 non-null  int64 
 4   count_floors_pre_eq                     260601 non-null  int64 
 5   age                                     260601 non-null  int64 
 6   area_percentage                         260601 non-null  int64 
 7   height_percentage                       260601 non-null  int64 
 8   land_surface_condition                  260601 non-null  object
 9   foundation_type                         260601 non-null  object
 10  roof_type                               260601 non-null  object
 11  ground_floor_type                       260601 non-null  object
 12  other_floor_type                        260601 non-null  object
 13  position                                260601 non-null  object
 14  plan_configuration                      260601 non-null  object
 15  has_superstructure_adobe_mud            260601 non-null  int64 
 16  has_superstructure_mud_mortar_stone     260601 non-null  int64 
 17  has_superstructure_stone_flag           260601 non-null  int64 
 18  has_superstructure_cement_mortar_stone  260601 non-null  int64 
 19  has_superstructure_mud_mortar_brick     260601 non-null  int64 
 20  has_superstructure_cement_mortar_brick  260601 non-null  int64 
 21  has_superstructure_timber               260601 non-null  int64 
 22  has_superstructure_bamboo               260601 non-null  int64 
 23  has_superstructure_rc_non_engineered    260601 non-null  int64 
 24  has_superstructure_rc_engineered        260601 non-null  int64 
 25  has_superstructure_other                260601 non-null  int64 
 26  legal_ownership_status                  260601 non-null  object
 27  count_families                          260601 non-null  int64 
 28  has_secondary_use                       260601 non-null  int64 
 29  has_secondary_use_agriculture           260601 non-null  int64 
 30  has_secondary_use_hotel                 260601 non-null  int64 
 31  has_secondary_use_rental                260601 non-null  int64 
 32  has_secondary_use_institution           260601 non-null  int64 
 33  has_secondary_use_school                260601 non-null  int64 
 34  has_secondary_use_industry              260601 non-null  int64 
 35  has_secondary_use_health_post           260601 non-null  int64 
 36  has_secondary_use_gov_office            260601 non-null  int64 
 37  has_secondary_use_use_police            260601 non-null  int64 
 38  has_secondary_use_other                 260601 non-null  int64 
 39  damage_grade                            260601 non-null  int64 
dtypes: int64(32), object(8)
memory usage: 81.5+ MB


train.select_dtypes(include=['int64']).columns

Index(['building_id', 'geo_level_1_id', 'geo_level_2_id', 'geo_level_3_id',
       'count_floors_pre_eq', 'age', 'area_percentage', 'height_percentage',
       'has_superstructure_adobe_mud', 'has_superstructure_mud_mortar_stone',
       'has_superstructure_stone_flag',
       'has_superstructure_cement_mortar_stone',
       'has_superstructure_mud_mortar_brick',
       'has_superstructure_cement_mortar_brick', 'has_superstructure_timber',
       'has_superstructure_bamboo', 'has_superstructure_rc_non_engineered',
       'has_superstructure_rc_engineered', 'has_superstructure_other',
       'count_families', 'has_secondary_use', 'has_secondary_use_agriculture',
       'has_secondary_use_hotel', 'has_secondary_use_rental',
       'has_secondary_use_institution', 'has_secondary_use_school',
       'has_secondary_use_industry', 'has_secondary_use_health_post',
       'has_secondary_use_gov_office', 'has_secondary_use_use_police',
       'has_secondary_use_other', 'damage_grade'],
      dtype='object')


train.select_dtypes(include=['object']).columns

Index(['land_surface_condition', 'foundation_type', 'roof_type',
       'ground_floor_type', 'other_floor_type', 'position',
       'plan_configuration', 'legal_ownership_status'],
      dtype='object')


train.describe()


train.describe(include='all')


train['damage_grade'] = train['damage_grade'].astype('category')


train['damage_grade'] = train['damage_grade'].replace({1:'Low', 2:'Medium', 3:'High'})


train['damage_grade'].unique()

array(['High', 'Medium', 'Low'], dtype=object)


train['damage_grade'].value_counts()

Medium    148259
High       87218
Low        25124
Name: damage_grade, dtype: int64


plt.figure(figsize=(7,7))

sns.set_theme()
sns.countplot(train['damage_grade'], order=['Low', 'Medium', 'High'])
plt.xlabel('Damage Grade', fontsize=12)
plt.ylabel('Frequency', fontsize=12)
plt.title('Count of Damage Grade', fontsize=15)
plt.tight_layout()
plt.show()


print('Jumlah Bangunan dengan tingkat kerusakan rendah : {}'.format(sum(train['damage_grade']=='Low')))
print('Jumlah Bangunan dengan tingkat kerusakan menengah : {}'.format(sum(train['damage_grade']=='Medium')))
print('Jumlah Bangunan dengan tingkat kerusakan tinggi : {}'.format(sum(train['damage_grade']=='High')))
print('')
print('Persentase Bangunan dengan tingkat kerusakan rendah : {:.2f}%'.format(sum(train['damage_grade']=='Low')/train.shape[0]*100))
print('Persentase Bangunan dengan tingkat kerusakan menengah : {:.2f}%'.format(sum(train['damage_grade']=='Medium')/train.shape[0]*100))
print('Persentase Bangunan dengan tingkat kerusakan tinggi : {:.2f}%'.format(sum(train['damage_grade']=='High')/train.shape[0]*100))

Jumlah Bangunan dengan tingkat kerusakan rendah : 25124
Jumlah Bangunan dengan tingkat kerusakan menengah : 148259
Jumlah Bangunan dengan tingkat kerusakan tinggi : 87218

Persentase Bangunan dengan tingkat kerusakan rendah : 9.64%
Persentase Bangunan dengan tingkat kerusakan menengah : 56.89%
Persentase Bangunan dengan tingkat kerusakan tinggi : 33.47%


def plot_box(data, col_y, y_label, title, col_x = 'damage_grade', x_label = 'Damage Grade'):
    plt.figure(figsize=(7,7))
    sns.set_theme()
    sns.boxplot(data[col_x], data[col_y], order=['Low', 'Medium', 'High'])
    plt.xlabel(x_label, fontsize=12)
    plt.ylabel(y_label, fontsize=12)
    plt.title(title, fontsize=14)
    plt.show()


plot_box(train, 'geo_level_1_id', 'Geo Level 1','Damage Grade Per Geographic Region for Geo Level 1')


plot_box(train, 'geo_level_2_id', 'Geo Level 2','Damage Grade Per Geographic Region for Geo Level 2')


plot_box(train, 'geo_level_3_id', 'Geo Level 3','Damage Grade Per Geographic Region for Geo Level 3')


train['count_floors_pre_eq'].value_counts()

2    156623
3     55617
1     40441
4      5424
5      2246
6       209
7        39
8         1
9         1
Name: count_floors_pre_eq, dtype: int64


plt.figure(figsize=(12,8))

sns.countplot(train['count_floors_pre_eq'], hue=train['damage_grade'], hue_order=['Low', 'Medium', 'High'])
plt.xlabel('Number of Floor', fontsize=12)
plt.ylabel('Frequency', fontsize=12)
plt.title('Number of Floor Before Earthquake', fontsize=15)
plt.legend(loc='upper right')
plt.show()


plt.figure(figsize=(20,8))

sns.countplot(train['age'], hue=train['damage_grade'], hue_order=['Low', 'Medium', 'High'])
plt.xlabel('Age', fontsize=12)
plt.ylabel('Frequency', fontsize=12)
plt.title('Age of Building', fontsize=15)
plt.legend(loc='upper right')
plt.show()


plt.figure(figsize=(20,8))

segment_1_age = train[train['age']<=50]
sns.countplot(segment_1_age['age'], hue=segment_1_age['damage_grade'], hue_order=['Low', 'Medium', 'High'])
plt.xlabel('Age', fontsize=12)
plt.ylabel('Frequency', fontsize=12)
plt.title('Damage grade to building with less than equal 50 years old', fontsize=15)
plt.legend(loc='upper right')
plt.show()


plt.figure(figsize=(20,8))

segment_2_age = train[(train['age']>50) & (train['age']<=100)]
sns.countplot(segment_2_age['age'], hue=segment_2_age['damage_grade'], hue_order=['Low', 'Medium', 'High'])
plt.xlabel('Age', fontsize=12)
plt.ylabel('Frequency', fontsize=12)
plt.title('Damage grade to building with age from 51 to 100 years old', fontsize=15)
plt.legend(loc='upper right')
plt.show()


plt.figure(figsize=(20,8))

segment_3_age = train[(train['age']>100) & (train['age']<=150)]
sns.countplot(segment_3_age['age'], hue=segment_3_age['damage_grade'], hue_order=['Low', 'Medium', 'High'])
plt.xlabel('Age', fontsize=12)
plt.ylabel('Frequency', fontsize=12)
plt.title('Damage grade to building with age from 101 to 150 years old', fontsize=15)
plt.legend(loc='upper left')
plt.show()


plt.figure(figsize=(20,8))

segment_4_age = train[train['age']>150]
sns.countplot(segment_4_age['age'], hue=segment_4_age['damage_grade'], hue_order=['Low', 'Medium', 'High'])
plt.xlabel('Age', fontsize=12)
plt.ylabel('Frequency', fontsize=12)
plt.title('Damage grade to building with greater than 150 years old', fontsize=15)
plt.legend(loc='upper left')
plt.show()


train['area_percentage'].unique()

array([  6,   8,   5,   9,   3,  13,   7,   4,  12,  16,  11,  27,  10,
        15,  14,  17,  21,  37,  19,   2,  28,  38,  56,   1,  20,  24,
        34,  26,  18,  31,  25,  23,  22,  32,  47,  36,  40,  29,  42,
        55,  35,  39, 100,  50,  51,  43,  30,  62,  85,  33,  45,  52,
        57,  49,  67,  66,  54,  75,  65,  58,  48,  64,  63,  46,  59,
        86,  78,  41,  44,  61,  70,  77,  73,  72,  53,  60,  84,  83,
        76,  96,  80,  90,  82,  69], dtype=int64)


train['area_percentage'].value_counts()

6      42013
7      36752
5      32724
8      28445
9      22199
       ...  
55         1
80         1
90         1
82         1
100        1
Name: area_percentage, Length: 84, dtype: int64


plt.figure(figsize=(20,8))

sns.countplot(train['area_percentage'], hue=train['damage_grade'], hue_order=['Low', 'Medium', 'High'])
plt.xlabel('Area Percentage', fontsize=12)
plt.xticks(rotation=45)
plt.ylabel('Frequency', fontsize=12)
plt.title('Damage distribution based on Area Percentage', fontsize=15)
plt.legend(loc='upper right')
plt.show()


train['height_percentage'].unique()

array([ 5,  7,  9,  4,  6,  3, 10,  8,  2, 32, 12, 11, 13, 16, 15, 18, 26,
       17, 19, 14, 20, 23, 21, 25, 24, 28, 31], dtype=int64)


train['height_percentage'].value_counts()

5     78513
6     46477
4     37763
7     35465
3     25957
8     13902
2      9305
9      5376
10     4492
11      917
12      907
13      759
15      292
16      179
32       75
18       71
14       66
20       33
21       13
23       11
17        9
19        7
24        4
25        3
26        2
28        2
31        1
Name: height_percentage, dtype: int64


plt.figure(figsize=(20,8))

sns.countplot(train['height_percentage'], hue=train['damage_grade'], hue_order=['Low', 'Medium', 'High'])
plt.xlabel('Height Percentage', fontsize=12)
plt.ylabel('Frequency', fontsize=12)
plt.title('Damage distribution based on Height Percentage', fontsize=15)
plt.legend(loc='upper right')
plt.show()


train['count_families'].value_counts()

1    226115
0     20862
2     11294
3      1802
4       389
5       104
6        22
7         7
9         4
8         2
Name: count_families, dtype: int64


plt.figure(figsize=(20,8))

sns.countplot(train['count_families'], hue=train['damage_grade'], hue_order=['Low', 'Medium', 'High'])
plt.xlabel('Count Families', fontsize=12)
plt.ylabel('Frequency', fontsize=12)
plt.title('Damage distribution based on Number of Family', fontsize=15)
plt.legend(loc='upper right')
plt.show()


def cat_plot(data, col_x, x_label, col_y = 'damage_grade'):
    plt.figure(figsize=(10,4))

    sns.countplot(data[col_x], hue=data[col_y], hue_order=['Low', 'Medium', 'High'])
    plt.xlabel(x_label, fontsize=12)
    plt.ylabel('Frequency', fontsize=12)
    plt.title('Damage distribution based on {}'.format(x_label), fontsize=15)
    plt.legend(loc='upper right')
    plt.show()


train['land_surface_condition'].value_counts()

t    216757
n     35528
o      8316
Name: land_surface_condition, dtype: int64


cat_plot(train, 'land_surface_condition', 'Land Surface Condition')


train['foundation_type'].value_counts()

r    219196
w     15118
u     14260
i     10579
h      1448
Name: foundation_type, dtype: int64


cat_plot(train, 'foundation_type', 'Foundation Type')


train['roof_type'].value_counts()

n    182842
q     61576
x     16183
Name: roof_type, dtype: int64


cat_plot(train, 'roof_type', 'Roof Type')


train['ground_floor_type'].value_counts()

f    209619
x     24877
v     24593
z      1004
m       508
Name: ground_floor_type, dtype: int64


cat_plot(train, 'ground_floor_type', 'Ground Floor Type')


train['other_floor_type'].value_counts()

q    165282
x     43448
j     39843
s     12028
Name: other_floor_type, dtype: int64


cat_plot(train, 'other_floor_type', 'Other Floor Type')


train['position'].value_counts()

s    202090
t     42896
j     13282
o      2333
Name: position, dtype: int64


cat_plot(train, 'position', 'Position')


train['plan_configuration'].value_counts()

d    250072
q      5692
u      3649
s       346
c       325
a       252
o       159
m        46
n        38
f        22
Name: plan_configuration, dtype: int64


cat_plot(train, 'plan_configuration', 'Plan Configuration')


train['legal_ownership_status'].value_counts()

v    250939
a      5512
w      2677
r      1473
Name: legal_ownership_status, dtype: int64


cat_plot(train, 'legal_ownership_status', 'Legal Ownership Status')


def count_plot(data, col_x, x_label, title, col_y = 'damage_grade'):
    plt.figure(figsize=(10,4))

    p = sns.countplot(data[col_x], hue=data[col_y], hue_order=['Low', 'Medium', 'High'])
    plt.xlabel(x_label, fontsize=12)
    p.set_xticklabels(['No', 'Yes'])
    plt.ylabel('Frequency', fontsize=12)
    plt.title('Damage distribution that has {}'.format(title), fontsize=15)
    plt.legend(loc='upper right')
    plt.show()


train['has_superstructure_adobe_mud'].value_counts()

0    237500
1     23101
Name: has_superstructure_adobe_mud, dtype: int64


count_plot(train, 'has_superstructure_adobe_mud', 'Has Superstructure Adobe Mud', 'Superstructure Adobe Mud')


train['has_superstructure_mud_mortar_stone'].value_counts()

1    198561
0     62040
Name: has_superstructure_mud_mortar_stone, dtype: int64


count_plot(train, 'has_superstructure_mud_mortar_stone', 'Has Superstructure Mud Mortar Stone', 'Superstructure Mud Mortar Stone')


train['has_superstructure_stone_flag'].value_counts()

0    251654
1      8947
Name: has_superstructure_stone_flag, dtype: int64


count_plot(train, 'has_superstructure_stone_flag', 'Has Superstructure Stone', 'Superstructure Stone')


train['has_superstructure_cement_mortar_stone'].value_counts()

0    255849
1      4752
Name: has_superstructure_cement_mortar_stone, dtype: int64


count_plot(train, 'has_superstructure_cement_mortar_stone', 'Has Superstructure Cement Mortar Stone', 'Superstructure Cement Mortar Stone')


train['has_superstructure_mud_mortar_brick'].value_counts()

0    242840
1     17761
Name: has_superstructure_mud_mortar_brick, dtype: int64


count_plot(train, 'has_superstructure_mud_mortar_brick', 'Has Superstructure Cement Mud Mortar Brick', 'Superstructure Cement Mud Mortar Brick')


train['has_superstructure_cement_mortar_brick'].value_counts()

0    240986
1     19615
Name: has_superstructure_cement_mortar_brick, dtype: int64


count_plot(train, 'has_superstructure_cement_mortar_brick', 'Has Superstructure Cement Mortar Brick', 'Superstructure Cement Mortar Brick')


train['has_superstructure_timber'].value_counts()

0    194151
1     66450
Name: has_superstructure_timber, dtype: int64


count_plot(train, 'has_superstructure_timber', 'Has Superstructure Timber', 'Superstructure Timber')


train['has_superstructure_bamboo'].value_counts()

0    238447
1     22154
Name: has_superstructure_bamboo, dtype: int64


count_plot(train, 'has_superstructure_bamboo', 'Has Superstructure Bamboo', 'Superstructure Bamboo')


train['has_superstructure_rc_non_engineered'].value_counts()

0    249502
1     11099
Name: has_superstructure_rc_non_engineered, dtype: int64


count_plot(train, 'has_superstructure_rc_non_engineered', 'Has Superstructure RC Non Engineered', 'Superstructure RC Non Engineered')


train['has_superstructure_rc_engineered'].value_counts()

0    256468
1      4133
Name: has_superstructure_rc_engineered, dtype: int64


count_plot(train, 'has_superstructure_rc_engineered', 'Has Superstructure RC Engineered', 'Superstructure RC Engineered')


train['has_superstructure_other'].value_counts()

0    256696
1      3905
Name: has_superstructure_other, dtype: int64


count_plot(train, 'has_superstructure_other', 'Has Superstructure Other', 'Superstructure Other')


train['has_secondary_use'].value_counts()

0    231445
1     29156
Name: has_secondary_use, dtype: int64


count_plot(train, 'has_secondary_use', 'Has Secondary Use', 'Secondary Use')


train['has_secondary_use_agriculture'].value_counts()

0    243824
1     16777
Name: has_secondary_use_agriculture, dtype: int64


count_plot(train, 'has_secondary_use_agriculture', 'Has Secondary Use Agriculture', 'Secondary Use Agriculture')


train['has_secondary_use_hotel'].value_counts()

0    251838
1      8763
Name: has_secondary_use_hotel, dtype: int64


count_plot(train, 'has_secondary_use_hotel', 'Has Secondary Use Hotel', 'Secondary Use Hotel')


train['has_secondary_use_rental'].value_counts()

0    258490
1      2111
Name: has_secondary_use_rental, dtype: int64


count_plot(train, 'has_secondary_use_rental', 'Has Secondary Use Rental', 'Secondary Use Rental')


train['has_secondary_use_institution'].value_counts()

0    260356
1       245
Name: has_secondary_use_institution, dtype: int64


count_plot(train, 'has_secondary_use_institution', 'Has Secondary Use Institution', 'Secondary Use Institution')


train['has_secondary_use_school'].value_counts()

0    260507
1        94
Name: has_secondary_use_school, dtype: int64


count_plot(train, 'has_secondary_use_school', 'Has Secondary Use School', 'Secondary Use School')


train['has_secondary_use_industry'].value_counts()

0    260322
1       279
Name: has_secondary_use_industry, dtype: int64


count_plot(train, 'has_secondary_use_industry', 'Has Secondary Use Industry', 'Secondary Use Industry')


train['has_secondary_use_health_post'].value_counts()

0    260552
1        49
Name: has_secondary_use_health_post, dtype: int64


count_plot(train, 'has_secondary_use_health_post', 'Has Secondary Use Health Post', 'Secondary Use Health Post')


train['has_secondary_use_gov_office'].value_counts()

0    260563
1        38
Name: has_secondary_use_gov_office, dtype: int64


count_plot(train, 'has_secondary_use_gov_office', 'Has Secondary Use Goverment Office', 'Secondary Use Goverment Office')


train['has_secondary_use_use_police'].value_counts()

0    260578
1        23
Name: has_secondary_use_use_police, dtype: int64


count_plot(train, 'has_secondary_use_use_police', 'Has Secondary Use Police', 'Secondary Use Police')


train['has_secondary_use_other'].value_counts()

0    259267
1      1334
Name: has_secondary_use_other, dtype: int64


count_plot(train, 'has_secondary_use_other', 'Has Secondary Use Other', 'Secondary Use Other')


def facet_plot(data, col, x, x_label):
    grid = sns.FacetGrid(data, col=col, row='damage_grade', size=4, aspect=2, sharey=False)
    grid.map(plt.hist, x, bins=20)
    grid.set_xlabels(x_label)
    grid.set_ylabels('Frequency')
    plt.show()


facet_plot(train, 'has_superstructure_rc_engineered', 'count_floors_pre_eq', 'Number of Floor')


facet_plot(train, 'has_superstructure_rc_non_engineered', 'count_floors_pre_eq', 'Number of Floor')


facet_plot(train, 'has_superstructure_mud_mortar_stone', 'count_floors_pre_eq', 'Number of Floor')


facet_plot(train, 'has_superstructure_cement_mortar_brick', 'count_floors_pre_eq', 'Number of Floor')


facet_plot(train, 'has_superstructure_rc_engineered', 'height_percentage', 'Height')


facet_plot(train, 'has_superstructure_rc_non_engineered', 'height_percentage', 'Height')


facet_plot(train, 'has_superstructure_mud_mortar_stone', 'height_percentage', 'Height')


facet_plot(train, 'has_superstructure_cement_mortar_brick', 'height_percentage', 'Height')


facet_plot(train, 'has_superstructure_rc_engineered', 'area_percentage', 'Area')


facet_plot(train, 'has_superstructure_rc_non_engineered', 'area_percentage', 'Area')


facet_plot(train, 'has_superstructure_mud_mortar_stone', 'area_percentage', 'Area')


facet_plot(train, 'has_superstructure_cement_mortar_brick', 'area_percentage', 'Area')


train.columns

Index(['building_id', 'geo_level_1_id', 'geo_level_2_id', 'geo_level_3_id',
       'count_floors_pre_eq', 'age', 'area_percentage', 'height_percentage',
       'land_surface_condition', 'foundation_type', 'roof_type',
       'ground_floor_type', 'other_floor_type', 'position',
       'plan_configuration', 'has_superstructure_adobe_mud',
       'has_superstructure_mud_mortar_stone', 'has_superstructure_stone_flag',
       'has_superstructure_cement_mortar_stone',
       'has_superstructure_mud_mortar_brick',
       'has_superstructure_cement_mortar_brick', 'has_superstructure_timber',
       'has_superstructure_bamboo', 'has_superstructure_rc_non_engineered',
       'has_superstructure_rc_engineered', 'has_superstructure_other',
       'legal_ownership_status', 'count_families', 'has_secondary_use',
       'has_secondary_use_agriculture', 'has_secondary_use_hotel',
       'has_secondary_use_rental', 'has_secondary_use_institution',
       'has_secondary_use_school', 'has_secondary_use_industry',
       'has_secondary_use_health_post', 'has_secondary_use_gov_office',
       'has_secondary_use_use_police', 'has_secondary_use_other',
       'damage_grade'],
      dtype='object')


features = train.copy()


features['damage_grade'] = features['damage_grade'].replace({'Low':1, 'Medium':2, 'High':3})
features['damage_grade'] = features['damage_grade'].astype('int64')


corr_matrix = features.corr()
corr_matrix


plt.figure(figsize=(20,20))

sns.heatmap(corr_matrix, annot=True, cmap='viridis', square=True, fmt='.2f', annot_kws={'size': 10})
plt.show()


plt.figure(figsize=(16,16))

corr_matrix['damage_grade'].drop('damage_grade').sort_values(ascending=False).plot(kind='barh')
plt.show()


list_corr = corr_matrix['damage_grade'].drop('damage_grade')
list_corr

building_id                               0.001063
geo_level_1_id                           -0.072347
geo_level_2_id                            0.043161
geo_level_3_id                            0.007932
count_floors_pre_eq                       0.122308
age                                       0.029273
area_percentage                          -0.125221
height_percentage                         0.048130
has_superstructure_adobe_mud              0.055314
has_superstructure_mud_mortar_stone       0.291325
has_superstructure_stone_flag             0.066039
has_superstructure_cement_mortar_stone   -0.060295
has_superstructure_mud_mortar_brick       0.014561
has_superstructure_cement_mortar_brick   -0.254131
has_superstructure_timber                -0.069852
has_superstructure_bamboo                -0.063051
has_superstructure_rc_non_engineered     -0.158145
has_superstructure_rc_engineered         -0.179014
has_superstructure_other                 -0.030224
count_families                            0.056151
has_secondary_use                        -0.079630
has_secondary_use_agriculture             0.011309
has_secondary_use_hotel                  -0.097942
has_secondary_use_rental                 -0.083754
has_secondary_use_institution            -0.028728
has_secondary_use_school                 -0.011692
has_secondary_use_industry               -0.011024
has_secondary_use_health_post            -0.008543
has_secondary_use_gov_office             -0.009378
has_secondary_use_use_police             -0.001656
has_secondary_use_other                  -0.016334
Name: damage_grade, dtype: float64


list_corr.loc[[
              'has_superstructure_adobe_mud', 
              'has_superstructure_mud_mortar_stone',
              'has_superstructure_stone_flag',
              'has_superstructure_cement_mortar_stone',
              'has_superstructure_mud_mortar_brick',
              'has_superstructure_cement_mortar_brick', 
              'has_superstructure_timber',
              'has_superstructure_bamboo', 
              'has_superstructure_rc_non_engineered',
              'has_superstructure_rc_engineered', 
              'has_superstructure_other'
              ]]

has_superstructure_adobe_mud              0.055314
has_superstructure_mud_mortar_stone       0.291325
has_superstructure_stone_flag             0.066039
has_superstructure_cement_mortar_stone   -0.060295
has_superstructure_mud_mortar_brick       0.014561
has_superstructure_cement_mortar_brick   -0.254131
has_superstructure_timber                -0.069852
has_superstructure_bamboo                -0.063051
has_superstructure_rc_non_engineered     -0.158145
has_superstructure_rc_engineered         -0.179014
has_superstructure_other                 -0.030224
Name: damage_grade, dtype: float64


list_corr.loc[['count_floors_pre_eq', 'age', 'area_percentage', 'height_percentage']]

count_floors_pre_eq    0.122308
age                    0.029273
area_percentage       -0.125221
height_percentage      0.048130
Name: damage_grade, dtype: float64


df = train.copy()
df = df.set_index('building_id')


df_num = df.select_dtypes(include=['int64', 'float64'])
df_cat = df.select_dtypes(include=['object'])


print('Korelasi feature height_percentange dengan damage_grade : {:.4f}'.format(target_corr_matrix['height_percentage']))
print('Korelasi feature has_secondary_use_agriculture dengan damage_grade : {:.4f}'.format(target_corr_matrix['has_secondary_use_agriculture']))

Korelasi feature height_percentange dengan damage_grade : 0.0481
Korelasi feature has_secondary_use_agriculture dengan damage_grade : 0.0113


df_num.head()


df_num = df_num.drop('has_secondary_use_agriculture', axis=1)


df_num.shape

(260601, 30)


df_cat.head()


df_cat = pd.get_dummies(df_cat)


df_cat.head()


df_cat.shape

(260601, 38)


category_rank_feature = SelectKBest(score_func=chi2, k=38)
rank_feature = category_rank_feature.fit(df_cat, df_num['damage_grade'])
rank_feature.scores_

array([3.14202700e+02, 6.77046429e+01, 6.77640172e+01, 4.15860728e+02,
       2.79293047e+04, 6.39195232e+03, 5.49424844e+03, 8.31579458e+03,
       1.04584801e+03, 1.15697625e+03, 2.80485950e+04, 3.68489235e+03,
       9.75543604e+01, 3.24654211e+04, 6.05516649e+01, 1.22429528e+02,
       7.42291993e+03, 5.10846128e+03, 1.85494082e+04, 4.06602490e+02,
       2.59354872e+02, 1.48779877e+02, 4.93714399e+01, 6.22421247e+02,
       1.09200110e+02, 1.50572497e+02, 3.04753264e+01, 3.34305427e+00,
       1.36435198e+01, 1.65187836e+00, 5.72019260e+01, 3.23110604e+02,
       3.31565900e+01, 1.01176162e+03, 2.26362245e+03, 5.34417037e+01,
       3.96056239e+01, 2.29493063e+02])


cat_score = pd.DataFrame(rank_feature.scores_, columns=['Score'])
cat_col = pd.DataFrame(df_cat.columns, columns=['Feature'])
univariate_cat_rank = pd.concat([cat_col, cat_score], axis=1)
univariate_cat_rank.set_index('Feature', inplace=True)
sorted_univariate_cat_rank = univariate_cat_rank.sort_values('Score', ascending=False)
#sorted_univariate_cat_rank


chi2_greater100 = sorted_univariate_cat_rank[sorted_univariate_cat_rank['Score']>100]


df_new = pd.concat([df_num, df_cat], axis=1)


df_new.sample(6)


data = df_new.copy()
y = data['damage_grade']
X = data[list(df_num.columns)+list(chi2_greater100.index)]
X.drop('damage_grade', axis=1, inplace=True)


X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)
X_train.shape, X_test.shape, y_train.shape, y_test.shape

((208480, 54), (52121, 54), (208480,), (52121,))


X_train.head()


X_test.head()


y_train.head().to_frame()


y_test.head().to_frame()


scaler = StandardScaler()

scaled_col = ['geo_level_1_id', 'geo_level_2_id', 'geo_level_3_id', 
              'count_floors_pre_eq', 'age', 'area_percentage', 
              'height_percentage']

X_train[scaled_col] = scaler.fit_transform(X_train[scaled_col])
X_test[scaled_col] = scaler.transform(X_test[scaled_col])


X_train.head()


X_test.head()


def make_confusion_matrix(pred, title):
    cam = confusion_matrix(y_test, pred)
    sns.heatmap(cam, annot=True, cmap='viridis', 
                xticklabels=['Low', 'Medium', 'High'],
                yticklabels=['Low', 'Medium', 'High'],
                fmt="d")
    plt.xlabel('Predicted Label')
    plt.ylabel('True Label')
    plt.title('Confusion Matrix Using {} Algorithm'.format(title))
    plt.show()


model_cv_1 = LogisticRegression(random_state=0)
kfold = KFold(n_splits=10)
scores_1 = cross_val_score(model_cv_1, X_train, y_train, scoring="f1_micro", cv=kfold)
result_1 = scores_1.mean()
print('Average F1 Score (Micro) for Logistic Algorithm is : {}'.format(result_1))

Average F1 Score (Micro) for Logistic Algorithm is : 0.5910878741366077


model_cv_2 = DecisionTreeClassifier(random_state=0)
kfold = KFold(n_splits=10)
scores_2 = cross_val_score(model_cv_2, X_train, y_train, scoring="f1_micro", cv=kfold)
result_2 = scores_2.mean()
print('Average F1 Score (Micro) for Decision Tree is : {}'.format(result_2))

Average F1 Score (Micro) for Decision Tree is : 0.6576026477359939


model_cv_3 = KNeighborsClassifier()
kfold = KFold(n_splits=10)
scores_3 = cross_val_score(model_cv_3, X_train, y_train, scoring="f1_micro", cv=kfold)
result_3 = scores_3.mean()
print('Average F1 Score (Micro) for K-Nearest Neighbors is : {}'.format(result_3))

Average F1 Score (Micro) for K-Nearest Neighbors is : 0.655962202609363


model_cv_4 = LinearDiscriminantAnalysis()
kfold = KFold(n_splits=10)
scores_4 = cross_val_score(model_cv_4, X_train, y_train, scoring="f1_micro", cv=kfold)
result_4 = scores_4.mean()
print('Average F1 Score (Micro) for Linear Discriminant Analysis is : {}'.format(result_4))

Average F1 Score (Micro) for Linear Discriminant Analysis is : 0.5867661166538757


model_cv_5 = GaussianNB()
kfold = KFold(n_splits=10)
scores_5 = cross_val_score(model_cv_5, X_train, y_train, scoring="f1_micro", cv=kfold)
result_5 = scores_5.mean()
print('Average F1 Score (Micro) for Naive Bayes is : {}'.format(result_5))

Average F1 Score (Micro) for Naive Bayes is : 0.430962202609363


model_cv_6 = XGBClassifier(verbosity=0)
kfold = KFold(n_splits=10)
scores_6 = cross_val_score(model_cv_6, X_train, y_train, scoring="f1_micro", cv=kfold)
result_6 = scores_6.mean()
print('Average F1 Score (Micro) for Extreme Gradient Boosting is : {}'.format(result_6))

Average F1 Score (Micro) for Extreme Gradient Boosting is : 0.7272831926323869


model_cv_7 = RandomForestClassifier(random_state=0)
kfold = KFold(n_splits=10)
scores_7 = cross_val_score(model_cv_7, X_train, y_train, scoring="f1_micro", cv=kfold)
result_7 = scores_7.mean()
print('Average F1 Score (Micro) for Random Forest is : {}'.format(result_7))

Average F1 Score (Micro) for Random Forest is : 0.7145577513430544


model_1 = LogisticRegression()
model_1.fit(X_train, y_train)
pred_1 = model_1.predict(X_test)
train_pred_1 = model_1.predict(X_train)

print(classification_report(y_test, pred_1))
make_confusion_matrix(pred_1, 'Logistic Regression')
train_score_1 = f1_score(y_train, train_pred_1, average='micro')
score_1 = f1_score(y_test, pred_1, average='micro')
print('Train F1 Score (Micro) for Logistic Regression Algorithm is : {}'.format(train_score_1))
print('Test F1 Score (Micro) for Logistic Regression Algorithm is : {}'.format(score_1))

              precision    recall  f1-score   support

           1       0.57      0.30      0.39      4979
           2       0.60      0.88      0.71     29703
           3       0.56      0.20      0.29     17439

    accuracy                           0.59     52121
   macro avg       0.58      0.46      0.47     52121
weighted avg       0.58      0.59      0.54     52121

Train F1 Score (Micro) for Logistic Regression Algorithm is : 0.591514773599386
Test F1 Score (Micro) for Logistic Regression Algorithm is : 0.5943669538189981


model_2 = DecisionTreeClassifier()
model_2.fit(X_train, y_train)
pred_2 = model_2.predict(X_test)
train_pred_2 = model_2.predict(X_train)

print(classification_report(y_test, pred_2))
make_confusion_matrix(pred_2, 'Decision Tree')
train_score_2 = f1_score(y_train, train_pred_2, average='micro')
score_2 = f1_score(y_test, pred_2, average='micro')
print('Train F1 Score (Micro) for Decision Tree Algorithm is : {}'.format(train_score_2))
print('Test F1 Score (Micro) for Decision Tree Algorithm is : {}'.format(score_2))

              precision    recall  f1-score   support

           1       0.48      0.52      0.50      4979
           2       0.71      0.71      0.71     29703
           3       0.62      0.61      0.62     17439

    accuracy                           0.66     52121
   macro avg       0.61      0.61      0.61     52121
weighted avg       0.66      0.66      0.66     52121

Train F1 Score (Micro) for Decision Tree Algorithm is : 0.9864207597851112
Test F1 Score (Micro) for Decision Tree Algorithm is : 0.6595614051917653


model_3 = KNeighborsClassifier()
model_3.fit(X_train, y_train)
pred_3 = model_3.predict(X_test)
train_pred_3 = model_3.predict(X_train)

print(classification_report(y_test, pred_3))
make_confusion_matrix(pred_3, 'K-Nearest Neighbors')
train_score_3 = f1_score(y_train, train_pred_3, average='micro')
score_3 = f1_score(y_test, pred_3, average='micro')
print('Train F1 Score (Micro) for K-Nearest Neighbors Algorithm is : {}'.format(train_score_3))
print('Test F1 Score (Micro) for K-Nearest Neighbors Algorithm is : {}'.format(score_3))

              precision    recall  f1-score   support

           1       0.52      0.46      0.49      4979
           2       0.69      0.75      0.72     29703
           3       0.63      0.56      0.59     17439

    accuracy                           0.66     52121
   macro avg       0.61      0.59      0.60     52121
weighted avg       0.65      0.66      0.66     52121

Train F1 Score (Micro) for K-Nearest Neighbors Algorithm is : 0.7662125863392172
Test F1 Score (Micro) for K-Nearest Neighbors Algorithm is : 0.65864047121122


model_4 = LinearDiscriminantAnalysis()
model_4.fit(X_train, y_train)
pred_4 = model_4.predict(X_test)
train_pred_4 = model_4.predict(X_train)

print(classification_report(y_test, pred_4))
make_confusion_matrix(pred_4, 'Linear Discriminant Analysis')
train_score_4 = f1_score(y_train, train_pred_4, average='micro')
score_4 = f1_score(y_test, pred_4, average='micro')
print('Train F1 Score (Micro) for Linear Discriminant Analysis Algorithm is : {}'.format(train_score_4))
print('Test F1 Score (Micro) for Linear Discriminant Analysis Algorithm is : {}'.format(score_4))

              precision    recall  f1-score   support

           1       0.49      0.50      0.50      4979
           2       0.60      0.85      0.70     29703
           3       0.57      0.18      0.27     17439

    accuracy                           0.59     52121
   macro avg       0.55      0.51      0.49     52121
weighted avg       0.58      0.59      0.54     52121

Train F1 Score (Micro) for Linear Discriminant Analysis Algorithm is : 0.587212202609363
Test F1 Score (Micro) for Linear Discriminant Analysis Algorithm is : 0.5883233245716698


model_5 = GaussianNB()
model_5.fit(X_train, y_train)
pred_5 = model_5.predict(X_test)
train_pred_5 = model_5.predict(X_train)

print(classification_report(y_test, pred_5))
make_confusion_matrix(pred_5, 'Naive Bayes')
train_score_5 = f1_score(y_train, train_pred_5, average='micro')
score_5 = f1_score(y_test, pred_5, average='micro')
print('Train F1 Score (Micro) for Naive Bayes Algorithm is : {}'.format(train_score_5))
print('Test F1 Score (Micro) for Naive Bayes Algorithm is : {}'.format(score_5))

              precision    recall  f1-score   support

           1       0.39      0.52      0.45      4979
           2       0.66      0.15      0.25     29703
           3       0.40      0.88      0.55     17439

    accuracy                           0.43     52121
   macro avg       0.49      0.52      0.42     52121
weighted avg       0.55      0.43      0.37     52121

Train F1 Score (Micro) for Naive Bayes Algorithm is : 0.43118284727551803
Test F1 Score (Micro) for Naive Bayes Algorithm is : 0.4337982770860114


model_6 = XGBClassifier(verbosity=0)
model_6.fit(X_train, y_train)
pred_6 = model_6.predict(X_test)
train_pred_6 = model_6.predict(X_train)

print(classification_report(y_test, pred_6))
make_confusion_matrix(pred_6, 'Extreme Gradient Boosting')
train_score_6 = f1_score(y_train, train_pred_6, average='micro')
score_6 = f1_score(y_test, pred_6, average='micro')
print('Train F1 Score (Micro) for Extreme Gradient Boosting Algorithm is : {}'.format(train_score_6))
print('Test F1 Score (Micro) for Extreme Gradient Boosting Algorithm is : {}'.format(score_6))

              precision    recall  f1-score   support

           1       0.69      0.45      0.54      4979
           2       0.72      0.86      0.78     29703
           3       0.75      0.58      0.66     17439

    accuracy                           0.73     52121
   macro avg       0.72      0.63      0.66     52121
weighted avg       0.73      0.73      0.72     52121

Train F1 Score (Micro) for Extreme Gradient Boosting Algorithm is : 0.7448196469685342
Test F1 Score (Micro) for Extreme Gradient Boosting Algorithm is : 0.7266936551485965


model_7 = RandomForestClassifier(random_state=0)
model_7.fit(X_train, y_train)
pred_7 = model_7.predict(X_test)
train_pred_7 = model_7.predict(X_train)

print(classification_report(y_test, pred_7))
make_confusion_matrix(pred_7, 'Random Forest')
train_score_7 = f1_score(y_train, train_pred_7, average='micro')
score_7 = f1_score(y_test, pred_7, average='micro')
print('Train F1 Score (Micro) for Random Forest Algorithm is : {}'.format(train_score_7))
print('Test F1 Score (Micro) for Random Forest Algorithm is : {}'.format(score_7))

              precision    recall  f1-score   support

           1       0.64      0.47      0.54      4979
           2       0.72      0.83      0.77     29703
           3       0.72      0.59      0.65     17439

    accuracy                           0.71     52121
   macro avg       0.69      0.63      0.65     52121
weighted avg       0.71      0.71      0.71     52121

Train F1 Score (Micro) for Random Forest Algorithm is : 0.9864111665387567
Test F1 Score (Micro) for Random Forest Algorithm is : 0.7142418602866407


names = ['Logistic Regression', 'Decision Tree', 'K-Nearest Neighbors', 'Linear Discriminant Analysis',
         'Naive Bayes', 'Extreme Gradient Boosting', 'Random Forest']
cv_score_f1 = [result_1, result_2, result_3, result_4, result_5, result_6, result_7]
train_score_f1 = [train_score_1, train_score_2, train_score_3, train_score_4, train_score_5, train_score_6, train_score_7]
score_f1 = [score_1, score_2, score_3, score_4, score_5, score_6, score_7]


compare = pd.DataFrame()
compare['Algorithm'] = names
compare['CV F1 Score'] = cv_score_f1
compare['F1 Score (Train)'] = train_score_f1
compare['F1 Score (Test)'] = score_f1
compare


xgb_params = {'n_estimators':range(50, 201, 50),
              'max_depth': range(3, 15)}
xgb_model = XGBClassifier(tree_method="gpu_hist", verbosity=0)
kfold = KFold(n_splits=4)
rscv_xgb = RandomizedSearchCV(estimator=xgb_model, param_distributions=xgb_params, n_iter=5,
                              scoring='f1_micro', cv=kfold, random_state=0, n_jobs=-1)

rscv_xgb.fit(X,y)
print("Optimal hyperparameter combination for XGBClassifier:", rscv_xgb.best_params_)
print()
print("Mean cross-validated training accuracy score:", rscv_xgb.best_score_)

Optimal hyperparameter combination for XGBClassifier: {'n_estimators': 150, 'max_depth': 10}

Mean cross-validated training accuracy score: 0.7340455365552411


params = {'n_estimators': range(50, 251, 50),
          'max_depth': range(15, 30),
          'max_features': ['sqrt', 'log2', None]}

rf_model = RandomForestClassifier(random_state=0)
kfold = KFold(n_splits=4)
rscv_rf = RandomizedSearchCV(estimator=rf_model, param_distributions=params, n_iter=5, 
                             scoring='f1_micro', cv=kfold, random_state=0, n_jobs=-1)

rscv_rf.fit(X,y)
print("Optimal hyperparameter combination for RandomForestClassifier:", rscv_rf.best_params_)
print()
print("Mean cross-validated training accuracy score:", rscv_rf.best_score_)

Optimal hyperparameter combination for RandomForestClassifier: {'n_estimators': 100, 'max_features': None, 'max_depth': 21}

Mean cross-validated training accuracy score: 0.7339687866382028


tuning_xgb_model = XGBClassifier(n_estimators=150, max_depth=10, verbosity=0, tree_method="gpu_hist")
tuning_xgb_model.fit(X_train,y_train)
pred_xgb = tuning_xgb_model.predict(X_test)
train_pred_xgb = tuning_xgb_model.predict(X_train)
print('F1 Score (Micro) on Training Data :', f1_score(y_train, train_pred_xgb, average='micro'))
print('F1 Score (Micro) on Testing Data :', f1_score(y_test, pred_xgb, average='micro'))

F1 Score (Micro) on Training Data : 0.7942680353031466
F1 Score (Micro) on Testing Data : 0.7356343892097237


tuning_rf_model = RandomForestClassifier(n_estimators=100, max_features=None, max_depth=21, random_state=0)
tuning_rf_model.fit(X_train,y_train)
pred_rf = tuning_rf_model.predict(X_test)
train_pred_rf = tuning_rf_model.predict(X_train)
print('F1 Score (Micro) on Training Data :', f1_score(y_train, train_pred_rf, average='micro'))
print('F1 Score (Micro) on Testing Data :', f1_score(y_test, pred_rf, average='micro'))

F1 Score (Micro) on Training Data : 0.8994675748273215
F1 Score (Micro) on Testing Data : 0.7350204332226934


fig, ax = plt.subplots(1,1,figsize=(10,10))
plot_importance(tuning_xgb_model, max_num_features=10, ax=ax)
plt.title('Top 10 Importance Feature')
plt.show()


features = X.columns
importances = tuning_rf_model.feature_importances_
indices = np.argsort(importances)
plt.figure(figsize=(10,10))
plt.title('Feature Importances')
plt.barh(range(len(indices[-10:])), importances[indices][-10:], color='b', align='center')
plt.yticks(range(len(indices[-10:])), features[indices][-10:])
plt.xlabel('Relative Importance')
plt.title('Top 10 Important Feature')
plt.show()


def final_confusion_matrix(pred, title):
    cam = confusion_matrix(y, pred)
    sns.heatmap(cam, annot=True, cmap='viridis', 
                xticklabels=['Low', 'Medium', 'High'],
                yticklabels=['Low', 'Medium', 'High'],
                fmt="d")
    plt.xlabel('Predicted Label')
    plt.ylabel('True Label')
    plt.title('Confusion Matrix Using {} Algorithm'.format(title))
    plt.show()


scaler_final = StandardScaler()
scaled_col = ['geo_level_1_id', 'geo_level_2_id', 'geo_level_3_id', 
              'count_floors_pre_eq', 'age', 'area_percentage', 
              'height_percentage']
X[scaled_col] = scaler_final.fit_transform(X[scaled_col])


final_model = XGBClassifier(n_estimators=150, max_depth=10, verbosity=0, tree_method="gpu_hist")
final_model.fit(X, y)
pred_final_model = final_model.predict(X)
print('F1 Score (Micro) on Training Data :', f1_score(y, pred_final_model, average='micro'))
print('Classification report: \n',classification_report(y, pred_final_model))
final_confusion_matrix(pred_final_model, 'Extreme Gradient Boosting')

F1 Score (Micro) on Training Data : 0.7867276027336809
Classification report: 
               precision    recall  f1-score   support

           1       0.83      0.60      0.70     25124
           2       0.77      0.89      0.83    148259
           3       0.81      0.66      0.73     87218

    accuracy                           0.79    260601
   macro avg       0.80      0.72      0.75    260601
weighted avg       0.79      0.79      0.78    260601


X.head()


test_data = test_value.copy()
test_data = test_data.set_index('building_id')
test_data.head()


test_data_num = test_data.select_dtypes(include=['int64', 'float64'])
test_data_cat = test_data.select_dtypes(include=['object'])


test_data_num = test_data_num.drop('has_secondary_use_agriculture', axis=1)


test_data_cat = pd.get_dummies(test_data_cat)


new_test_data = pd.concat([test_data_num, test_data_cat], axis=1)


scaler = StandardScaler()
scaled_test = new_test_data.copy()
data_train = df_new.copy()
scaled_col = ['geo_level_1_id', 'geo_level_2_id', 'geo_level_3_id', 
              'count_floors_pre_eq', 'age', 'area_percentage', 
              'height_percentage']

scaled_train = data_train[scaled_col]
scaler.fit(scaled_train)
feature_scaled_test = scaled_test[scaled_col]
feature_scaled_test = scaler.transform(feature_scaled_test)
scaled_test[scaled_col] = feature_scaled_test
data_test = scaled_test.copy()


data_test.head()


testing = data_test[list(df_num.drop('damage_grade', axis=1).columns)+list(chi2_greater100.index)]


pred_testing_data = final_model.predict(testing)


final = pd.DataFrame()
final['building_id'] = testing.index
final['damage_grade'] = pred_testing_data


final


final.to_csv('Damage Grade Prediction.csv', index=False)

	building_id	damage_grade
0	802906	3
1	28830	2
2	94947	3
3	590882	2
4	201944	3

	building_id	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	has_superstructure_adobe_mud	has_superstructure_mud_mortar_stone	...	has_secondary_use_hotel	has_secondary_use_rental	has_secondary_use_institution	has_secondary_use_school	has_secondary_use_industry	has_secondary_use_health_post	has_secondary_use_gov_office	has_secondary_use_use_police	has_secondary_use_other	damage_grade
count	2.606010e+05	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	...	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000
mean	5.256755e+05	13.900353	701.074685	6257.876148	2.129723	26.535029	8.018051	5.434365	0.088645	0.761935	...	0.033626	0.008101	0.000940	0.000361	0.001071	0.000188	0.000146	0.000088	0.005119	2.238272
std	3.045450e+05	8.033617	412.710734	3646.369645	0.727665	73.565937	4.392231	1.918418	0.284231	0.425900	...	0.180265	0.089638	0.030647	0.018989	0.032703	0.013711	0.012075	0.009394	0.071364	0.611814
min	4.000000e+00	0.000000	0.000000	0.000000	1.000000	0.000000	1.000000	2.000000	0.000000	0.000000	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	1.000000
25%	2.611900e+05	7.000000	350.000000	3073.000000	2.000000	10.000000	5.000000	4.000000	0.000000	1.000000	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	2.000000
50%	5.257570e+05	12.000000	702.000000	6270.000000	2.000000	15.000000	7.000000	5.000000	0.000000	1.000000	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	2.000000
75%	7.897620e+05	21.000000	1050.000000	9412.000000	2.000000	30.000000	9.000000	6.000000	0.000000	1.000000	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	3.000000
max	1.052934e+06	30.000000	1427.000000	12567.000000	9.000000	995.000000	100.000000	32.000000	1.000000	1.000000	...	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	3.000000

	building_id	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	land_surface_condition	foundation_type	...	has_secondary_use_hotel	has_secondary_use_rental	has_secondary_use_institution	has_secondary_use_school	has_secondary_use_industry	has_secondary_use_health_post	has_secondary_use_gov_office	has_secondary_use_use_police	has_secondary_use_other	damage_grade
count	2.606010e+05	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601	260601	...	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000	260601.000000
unique	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	3	5	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
top	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	t	r	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
freq	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	216757	219196	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
mean	5.256755e+05	13.900353	701.074685	6257.876148	2.129723	26.535029	8.018051	5.434365	NaN	NaN	...	0.033626	0.008101	0.000940	0.000361	0.001071	0.000188	0.000146	0.000088	0.005119	2.238272
std	3.045450e+05	8.033617	412.710734	3646.369645	0.727665	73.565937	4.392231	1.918418	NaN	NaN	...	0.180265	0.089638	0.030647	0.018989	0.032703	0.013711	0.012075	0.009394	0.071364	0.611814
min	4.000000e+00	0.000000	0.000000	0.000000	1.000000	0.000000	1.000000	2.000000	NaN	NaN	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	1.000000
25%	2.611900e+05	7.000000	350.000000	3073.000000	2.000000	10.000000	5.000000	4.000000	NaN	NaN	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	2.000000
50%	5.257570e+05	12.000000	702.000000	6270.000000	2.000000	15.000000	7.000000	5.000000	NaN	NaN	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	2.000000
75%	7.897620e+05	21.000000	1050.000000	9412.000000	2.000000	30.000000	9.000000	6.000000	NaN	NaN	...	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	3.000000
max	1.052934e+06	30.000000	1427.000000	12567.000000	9.000000	995.000000	100.000000	32.000000	NaN	NaN	...	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000	3.000000

	building_id	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	has_superstructure_adobe_mud	has_superstructure_mud_mortar_stone	...	has_secondary_use_hotel	has_secondary_use_rental	has_secondary_use_institution	has_secondary_use_school	has_secondary_use_industry	has_secondary_use_health_post	has_secondary_use_gov_office	has_secondary_use_use_police	has_secondary_use_other	damage_grade
building_id	1.000000	-0.002850	0.000347	-0.000393	-0.000654	-0.001476	-0.002070	0.000096	-0.000307	0.002423	...	0.001934	-0.002152	0.000706	-0.000362	0.002348	-0.000374	0.000538	-0.003116	-0.002295	0.001063
geo_level_1_id	-0.002850	1.000000	-0.061405	0.002718	-0.089364	-0.003908	0.071158	-0.063474	-0.018245	-0.152038	...	0.001911	0.023523	0.003700	0.002977	0.002655	-0.002303	0.001060	0.000523	-0.017992	-0.072347
geo_level_2_id	0.000347	-0.061405	1.000000	0.000921	0.047730	0.012594	-0.049443	0.035516	0.015833	0.076491	...	-0.008439	-0.030704	-0.004840	-0.004856	0.000687	-0.000757	-0.000152	0.001926	-0.013068	0.043161
geo_level_3_id	-0.000393	0.002718	0.000921	1.000000	-0.021646	-0.006385	-0.005643	-0.024507	-0.015732	0.026294	...	-0.002001	-0.007356	-0.007058	-0.004373	-0.000862	-0.002632	-0.000943	0.000269	-0.002463	0.007932
count_floors_pre_eq	-0.000654	-0.089364	0.047730	-0.021646	1.000000	0.086668	0.101071	0.772734	0.174852	-0.027116	...	0.077120	0.035425	0.016384	0.008833	-0.002611	0.006786	0.009639	0.003939	-0.002073	0.122308
age	-0.001476	-0.003908	0.012594	-0.006385	0.086668	1.000000	-0.004323	0.061074	0.068032	0.001321	...	-0.010021	0.001193	-0.004189	-0.003514	-0.003658	-0.002169	-0.001764	-0.001195	-0.004534	0.029273
area_percentage	-0.002070	0.071158	-0.049443	-0.005643	0.101071	-0.004323	1.000000	0.196645	0.026287	-0.225541	...	0.159885	0.105983	0.052212	0.050164	0.019421	0.015109	0.015290	0.004983	0.013111	-0.125221
height_percentage	0.000096	-0.063474	0.035516	-0.024507	0.772734	0.061074	0.196645	1.000000	0.149725	-0.106573	...	0.123551	0.068909	0.031366	0.020032	0.001946	0.011192	0.014660	0.004048	0.005397	0.048130
has_superstructure_adobe_mud	-0.000307	-0.018245	0.015833	-0.015732	0.174852	0.068032	0.026287	0.149725	1.000000	-0.306861	...	-0.012642	-0.003935	-0.004281	-0.002369	0.001762	-0.003292	-0.002648	-0.001493	-0.010074	0.055314
has_superstructure_mud_mortar_stone	0.002423	-0.152038	0.076491	0.026294	-0.027116	0.001321	-0.225541	-0.106573	-0.306861	1.000000	...	-0.159532	-0.117948	-0.036064	-0.023070	-0.025507	-0.008763	-0.011904	-0.003380	0.005628	0.291325
has_superstructure_stone_flag	0.001443	0.007677	-0.008743	0.019579	-0.041058	0.009129	-0.000473	-0.020260	0.006962	-0.034046	...	-0.008868	-0.011632	-0.000971	-0.001362	-0.004240	-0.002586	0.002959	0.000472	0.002127	0.066039
has_superstructure_cement_mortar_stone	-0.000827	0.023656	-0.009122	0.005966	-0.011329	-0.012128	0.073701	0.012960	-0.014554	-0.104288	...	0.072104	0.033757	0.007985	0.006473	0.006939	0.004406	0.012605	0.004825	0.014337	-0.060295
has_superstructure_mud_mortar_brick	-0.003768	-0.028854	0.014246	-0.023278	0.257279	0.079525	0.053362	0.209098	0.314979	-0.376208	...	0.025152	0.018368	-0.002334	-0.000326	0.011168	0.000733	-0.003266	-0.000920	-0.004890	0.014561
has_superstructure_cement_mortar_brick	0.001933	0.134121	-0.079875	-0.005116	-0.085821	-0.036992	0.210800	0.001698	-0.037549	-0.470715	...	0.139541	0.109545	0.032063	0.019858	0.026241	0.008817	0.007396	0.005061	0.000121	-0.254131
has_superstructure_timber	0.001967	0.144985	-0.023927	-0.003916	-0.056637	0.005855	-0.053965	-0.052402	0.011694	-0.040177	...	-0.027763	-0.026546	-0.005306	-0.003695	-0.000577	-0.004812	-0.004148	-0.002685	-0.014453	-0.069852
has_superstructure_bamboo	0.001266	0.074169	-0.032588	0.006143	-0.070487	-0.008374	-0.031641	-0.063342	0.011528	-0.054657	...	-0.030755	-0.019104	-0.004861	-0.003616	-0.003247	-0.004180	-0.002542	-0.002864	-0.008561	-0.063051
has_superstructure_rc_non_engineered	0.002472	0.016454	-0.038173	0.002265	0.007448	-0.025431	0.185047	0.086228	-0.036830	-0.222366	...	0.157997	0.102838	0.036314	0.021012	0.015176	0.008195	0.005322	0.002064	0.000582	-0.158145
has_superstructure_rc_engineered	-0.002259	0.015767	-0.036074	-0.012420	0.049000	-0.025552	0.224647	0.129698	-0.036782	-0.224509	...	0.139722	0.131416	0.050225	0.025087	0.005236	0.011700	0.031536	0.005347	0.008110	-0.179014
has_superstructure_other	0.002171	0.022670	-0.020144	-0.010604	-0.029888	-0.001520	0.007928	-0.019266	0.057545	-0.041705	...	0.007830	0.000130	-0.002753	0.000984	-0.001140	0.000612	-0.001489	-0.001159	0.006201	-0.030224
count_families	-0.000599	0.035108	-0.012439	-0.002479	0.086586	0.005309	0.088630	0.064316	0.033068	0.001576	...	-0.034615	0.022703	-0.022465	-0.020040	-0.021461	-0.010846	-0.004854	-0.014284	-0.016783	0.056151
has_secondary_use	0.002606	-0.010428	-0.001519	0.007375	0.052125	-0.008788	0.122401	0.091780	-0.013300	-0.087520	...	0.525564	0.254614	0.086429	0.053520	0.092237	0.038638	0.034025	0.026470	0.202099	-0.079630
has_secondary_use_agriculture	0.002631	-0.019786	0.017522	0.015995	-0.005294	-0.002194	-0.016478	-0.005390	-0.003917	0.058261	...	-0.048931	-0.023705	-0.008047	-0.004983	-0.008587	-0.003597	-0.003168	-0.002464	0.085034	0.011309
has_secondary_use_hotel	0.001934	0.001911	-0.008439	-0.002001	0.077120	-0.010021	0.159885	0.123551	-0.012642	-0.159532	...	1.000000	-0.016857	-0.005722	-0.003543	-0.006107	-0.002558	-0.002253	-0.001753	0.003920	-0.097942
has_secondary_use_rental	-0.002152	0.023523	-0.030704	-0.007356	0.035425	0.001193	0.105983	0.068909	-0.003935	-0.117948	...	-0.016857	1.000000	-0.002772	-0.001717	-0.002958	-0.001239	-0.001091	-0.000849	0.002516	-0.083754
has_secondary_use_institution	0.000706	0.003700	-0.004840	-0.007058	0.016384	-0.004189	0.052212	0.031366	-0.004281	-0.036064	...	-0.005722	-0.002772	1.000000	-0.000583	-0.001004	-0.000421	-0.000370	-0.000288	0.004818	-0.028728
has_secondary_use_school	-0.000362	0.002977	-0.004856	-0.004373	0.008833	-0.003514	0.050164	0.020032	-0.002369	-0.023070	...	-0.003543	-0.001717	-0.000583	1.000000	-0.000622	-0.000260	-0.000229	-0.000178	0.001469	-0.011692
has_secondary_use_industry	0.002348	0.002655	0.000687	-0.000862	-0.002611	-0.003658	0.019421	0.001946	0.001762	-0.025507	...	-0.006107	-0.002958	-0.001004	-0.000622	1.000000	-0.000449	-0.000395	-0.000308	0.004229	-0.011024
has_secondary_use_health_post	-0.000374	-0.002303	-0.000757	-0.002632	0.006786	-0.002169	0.015109	0.011192	-0.003292	-0.008763	...	-0.002558	-0.001239	-0.000421	-0.000260	-0.000449	1.000000	-0.000166	-0.000129	-0.000984	-0.008543
has_secondary_use_gov_office	0.000538	0.001060	-0.000152	-0.000943	0.009639	-0.001764	0.015290	0.014660	-0.002648	-0.011904	...	-0.002253	-0.001091	-0.000370	-0.000229	-0.000395	-0.000166	1.000000	-0.000113	0.003587	-0.009378
has_secondary_use_use_police	-0.003116	0.000523	0.001926	0.000269	0.003939	-0.001195	0.004983	0.004048	-0.001493	-0.003380	...	-0.001753	-0.000849	-0.000288	-0.000178	-0.000308	-0.000129	-0.000113	1.000000	-0.000674	-0.001656
has_secondary_use_other	-0.002295	-0.017992	-0.013068	-0.002463	-0.002073	-0.004534	0.013111	0.005397	-0.010074	0.005628	...	0.003920	0.002516	0.004818	0.001469	0.004229	-0.000984	0.003587	-0.000674	1.000000	-0.016334
damage_grade	0.001063	-0.072347	0.043161	0.007932	0.122308	0.029273	-0.125221	0.048130	0.055314	0.291325	...	-0.097942	-0.083754	-0.028728	-0.011692	-0.011024	-0.008543	-0.009378	-0.001656	-0.016334	1.000000

	land_surface_condition_n	land_surface_condition_o	land_surface_condition_t	foundation_type_h	foundation_type_i	foundation_type_r	foundation_type_u	foundation_type_w	roof_type_n	roof_type_q	...	plan_configuration_m	plan_configuration_n	plan_configuration_o	plan_configuration_q	plan_configuration_s	plan_configuration_u	legal_ownership_status_a	legal_ownership_status_r	legal_ownership_status_v	legal_ownership_status_w
building_id
802906	0	0	1	0	0	1	0	0	1	0	...	0	0	0	0	0	0	0	0	1	0
28830	0	1	0	0	0	1	0	0	1	0	...	0	0	0	0	0	0	0	0	1	0
94947	0	0	1	0	0	1	0	0	1	0	...	0	0	0	0	0	0	0	0	1	0
590882	0	0	1	0	0	1	0	0	1	0	...	0	0	0	0	0	0	0	0	1	0
201944	0	0	1	0	0	1	0	0	1	0	...	0	0	0	0	0	0	0	0	1	0

	building_id	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	land_surface_condition	foundation_type	...
0	802906	6	487	12198	2	30	6	5	t	r	...
1	28830	8	900	2812	2	10	8	7	o	r	...
2	94947	21	363	8973	2	10	5	5	t	r	...
3	590882	22	418	10694	2	10	6	5	t	r	...
4	201944	11	131	1488	3	30	8	9	t	r	...

	building_id	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	land_surface_condition	foundation_type	...	has_secondary_use_agriculture	has_secondary_use_rental
0	300051	17	596	11307	3	20	7	6	t	r	...	0	0
1	99355	6	141	11987	2	25	13	5	t	r	...	1	0
2	890251	22	19	10044	2	5	4	5	t	r	...	0	0
3	745817	26	39	633	1	0	19	3	t	r	...	0	1
4	421793	17	289	7970	3	15	8	7	t	r	...	0	0

	building_id	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	land_surface_condition	foundation_type	...	damage_grade
260596	688636	25	1335	1621	1	55	6	3	n	r	...	2
260597	669485	17	715	2060	2	0	6	5	t	r	...	3
260598	602512	17	51	8163	3	55	6	7	t	r	...	3
260599	151409	26	39	1851	2	10	14	6	t	r	...	2
260600	747594	21	9	9101	3	10	7	6	n	r	...	3

	building_id	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	land_surface_condition	foundation_type	...	damage_grade
23913	775524	8	659	7703	2	15	5	5	n	r	...	3
252595	85085	18	977	2847	1	20	6	3	t	r	...	3
72320	794653	17	414	11203	2	5	13	7	t	r	...	2
46911	756701	8	1225	3947	3	20	12	7	t	r	...	2
178042	616630	11	765	10814	2	30	6	6	t	r	...	3

	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	has_superstructure_adobe_mud	has_superstructure_mud_mortar_stone	has_superstructure_stone_flag	...	plan_configuration_m	plan_configuration_n	plan_configuration_o	plan_configuration_q	plan_configuration_s	plan_configuration_u	legal_ownership_status_a	legal_ownership_status_r	legal_ownership_status_v	legal_ownership_status_w
building_id
671063	8	900	2402	3	50	8	5	0	1	0	...	0	0	0	0	0	0	0	0	1	0
334036	3	1387	10110	5	30	7	11	1	0	0	...	0	0	0	0	0	0	0	0	1	0
855327	21	30	672	2	10	6	4	0	1	0	...	0	0	0	0	0	0	0	0	1	0
71688	12	234	6185	2	80	7	4	0	1	0	...	0	0	0	0	0	0	0	0	1	0
1026290	27	216	12323	2	10	4	5	1	0	0	...	0	0	0	0	0	0	0	0	0	1
981120	7	129	5248	2	5	9	4	0	1	0	...	0	0	0	0	0	0	0	0	1	0

	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	has_superstructure_adobe_mud	has_superstructure_mud_mortar_stone	has_superstructure_stone_flag	...	foundation_type_h	other_floor_type_x	plan_configuration_q	land_surface_condition_n	position_j	legal_ownership_status_w	plan_configuration_c	position_o	ground_floor_type_z	plan_configuration_a
building_id
882292	20	281	7097	2	15	6	7	0	1	0	...	0	0	0	0	0	0	0	0	0	0
193569	26	886	12157	2	10	6	6	0	0	0	...	0	0	0	0	0	0	0	0	0	0
898082	26	36	1125	1	0	5	3	0	0	0	...	0	0	0	1	0	0	0	0	0	0
136982	13	1372	4555	2	10	6	5	0	1	0	...	0	0	0	0	0	0	0	0	0	0
742018	8	206	6064	3	35	6	5	0	1	0	...	0	1	0	0	0	0	0	0	0	0

	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	has_superstructure_adobe_mud	has_superstructure_mud_mortar_stone	has_superstructure_stone_flag	...	foundation_type_h	other_floor_type_x	plan_configuration_q	land_surface_condition_n	position_j	legal_ownership_status_w	plan_configuration_c	position_o	ground_floor_type_z	plan_configuration_a
building_id
315409	6	449	484	2	0	9	8	0	1	0	...	0	0	0	0	0	0	0	0	0	0
838451	27	548	2513	2	0	5	5	0	1	0	...	0	0	0	0	0	0	0	0	0	0
808750	24	780	9035	3	25	4	7	0	1	0	...	0	0	0	0	0	0	0	0	0	0
322968	10	661	6346	3	30	6	7	0	1	0	...	0	0	0	0	0	0	0	0	0	0
212418	10	1207	276	2	15	5	5	0	1	0	...	0	0	0	0	0	0	0	0	0	0

	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	has_superstructure_adobe_mud	has_superstructure_mud_mortar_stone	has_superstructure_stone_flag	...	foundation_type_h	other_floor_type_x	plan_configuration_q	land_surface_condition_n	position_j	legal_ownership_status_w	plan_configuration_c	position_o	ground_floor_type_z	plan_configuration_a
building_id
882292	0.758641	-1.019753	0.228478	-0.178354	-0.156363	-0.461464	0.817733	0	1	0	...	0	0	0	0	0	0	0	0	0	0
193569	1.505643	0.446238	1.615688	-0.178354	-0.224906	-0.461464	0.295711	0	0	0	...	0	0	0	0	0	0	0	0	0	0
898082	1.505643	-1.613419	-1.408759	-1.556477	-0.361991	-0.690498	-1.270355	0	0	0	...	0	0	0	1	0	0	0	0	0	0
136982	-0.112861	1.623877	-0.468417	-0.178354	-0.224906	-0.461464	-0.226311	0	1	0	...	0	0	0	0	0	0	0	0	0	0
742018	-0.735362	-1.201487	-0.054721	1.199770	0.117809	-0.461464	-0.226311	0	1	0	...	0	1	0	0	0	0	0	0	0	0

	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	has_superstructure_adobe_mud	has_superstructure_mud_mortar_stone	has_superstructure_stone_flag	...	foundation_type_h	other_floor_type_x	plan_configuration_q	land_surface_condition_n	position_j	legal_ownership_status_w	plan_configuration_c	position_o	ground_floor_type_z	plan_configuration_a
building_id
315409	-0.984363	-0.612668	-1.584491	-0.178354	-0.361991	0.225640	1.339754	0	1	0	...	0	0	0	0	0	0	0	0	0	0
838451	1.630143	-0.372778	-1.028236	-0.178354	-0.361991	-0.690498	-0.226311	0	1	0	...	0	0	0	0	0	0	0	0	0	0
808750	1.256643	0.189387	0.759785	1.199770	-0.019277	-0.919532	0.817733	0	1	0	...	0	0	0	0	0	0	0	0	0	0
322968	-0.486362	-0.098965	0.022590	1.199770	0.049266	-0.461464	0.817733	0	1	0	...	0	0	0	0	0	0	0	0	0	0
212418	-0.486362	1.224061	-1.641514	-0.178354	-0.156363	-0.690498	-0.226311	0	1	0	...	0	0	0	0	0	0	0	0	0	0

	Algorithm	CV F1 Score	F1 Score (Train)	F1 Score (Test)
0	Logistic Regression	0.591088	0.591515	0.594367
1	Decision Tree	0.657603	0.986421	0.659561
2	K-Nearest Neighbors	0.655962	0.766213	0.658640
3	Linear Discriminant Analysis	0.586766	0.587212	0.588323
4	Naive Bayes	0.430962	0.431183	0.433798
5	Extreme Gradient Boosting	0.727283	0.744820	0.726694
6	Random Forest	0.714558	0.986411	0.714242

	building_id	damage_grade
0	300051	3
1	99355	2
2	890251	2
3	745817	1
4	421793	3
...	...	...
86863	310028	2
86864	663567	2
86865	1049160	2
86866	442785	2
86867	501372	1

	geo_level_1_id	geo_level_2_id	geo_level_3_id	count_floors_pre_eq	age	area_percentage	height_percentage	has_superstructure_adobe_mud	has_superstructure_mud_mortar_stone	has_superstructure_stone_flag	has_superstructure_cement_mortar_stone	has_superstructure_mud_mortar_brick	has_superstructure_cement_mortar_brick	has_superstructure_timber	has_superstructure_bamboo	has_superstructure_rc_non_engineered	has_superstructure_rc_engineered	has_superstructure_other	count_families	has_secondary_use	has_secondary_use_hotel	has_secondary_use_rental	has_secondary_use_institution	has_secondary_use_school	has_secondary_use_industry	has_secondary_use_health_post	has_secondary_use_gov_office	has_secondary_use_use_police	has_secondary_use_other	land_surface_condition_n	land_surface_condition_o	land_surface_condition_t	foundation_type_h	foundation_type_i	foundation_type_r	foundation_type_u	foundation_type_w	roof_type_n	roof_type_q	roof_type_x	ground_floor_type_f	ground_floor_type_m	ground_floor_type_v	ground_floor_type_x	ground_floor_type_z	other_floor_type_j	other_floor_type_q	other_floor_type_s	other_floor_type_x	position_j	position_o	position_s	position_t	plan_configuration_a	plan_configuration_c	plan_configuration_d	plan_configuration_f	plan_configuration_m	plan_configuration_n	plan_configuration_o	plan_configuration_q	plan_configuration_s	plan_configuration_u	legal_ownership_status_a	legal_ownership_status_r	legal_ownership_status_v	legal_ownership_status_w
building_id
300051	0.385835	-0.254597	1.384702	1.195989	-0.088832	-0.231785	0.294845	0	1	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	1	0	0	1	0	0	0	0	0	1	0	0	0	0	1	0	0	0	1	0	0	0	0	0	0	0	0	0	1	0
99355	-0.983414	-1.357066	1.571189	-0.178274	-0.020866	1.134266	-0.226419	0	1	0	0	0	0	0	0	0	0	0	1	1	0	0	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	1	0	0	1	0	0	0	0	0	1	0	0	0	0	1	0	0	0	1	0	0	0	0	0	0	0	0	0	1	0
890251	1.008221	-1.652673	1.038329	-0.178274	-0.292732	-0.914810	-0.226419	0	1	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	1	0	0	1	0	0	0	0	0	1	0	0	0	0	1	0	0	0	1	0	0	0	0	0	0	0	0	0	1	0
745817	1.506130	-1.604213	-1.542599	-1.552536	-0.360698	2.500317	-1.268946	0	0	0	0	0	1	0	0	0	0	0	2	1	0	1	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	0	0	1	0	0	1	0	0	1	0	0	0	0	0	0	1	0	0	1	0	0	0	0	0	0	0	0	0	1	0
421793	0.385835	-0.998461	0.469543	1.195989	-0.156799	-0.004110	0.816109	0	1	0	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	1	0	0	0	1	0	1	0	0	0	0	0	1	0	0	0	0	0	1	0	0	1	0	0	0	0	0	0	0	0	0	1	0

Predict Earthquake Damage to Building in Nepal¶

Reynara Ezra Pratama¶

Background¶

Business Understanding¶

Importing Library¶

Loading Dataset¶

Checking Data¶

Description¶

Data Preparation¶

Exploratory Data Analysis¶

Plot Label Data¶

Check Between Feature and Label¶

Univariate Analysis¶

Geo Level and Damage Grade¶

Number of Floor and Damage Grade¶

Age and Damage Grade¶

Area Percentage and Damage Grade¶

Height Percentage and Damage Grade¶

Count Families and Damage Grade¶

Categorical Feature and Damage Grade¶

Numerical Feature (Binary) and Damage Grade¶

Facetgrid¶

Superstructure RC, non RC, Mud Mortar Stone, Mud Mortar Brick, Cement Mortar Brick and Number of Floor¶

Superstructure RC, non RC, Mud Mortar Stone, Mud Mortar Brick, Cement Mortar Brick and Height Percentage¶

Superstructure RC, non RC, Mud Mortar Stone, Mud Mortar Brick, Cement Mortar Brick and Area Percentage¶

Multivariate Analysis¶

EDA (Recommendation Material and Characteristic)¶

Feature Selection¶

Split Data into Train Data and Validation Data¶

Scaling Feature¶

Building Model¶

Using Cross Validation¶

Logistic Regression¶

Decision Tree¶

K-Nearest Neighbors¶

Linear Discriminant Analysis¶

Naive Bayes¶

Extreme Gradient Boosting¶

Random Forest¶

Fit and Predict Several Algorithm¶

Logistic Regression¶

Decision Tree¶

K-Nearest Neighbors¶

Linear Discriminant Analysis¶

Naive Bayes¶

Extreme Gradient Boosting¶

Random Forest¶

Comparing Algorithm¶

Hyperparameter Tuning¶

Build Model with Parameter Tuning¶

Feature Importance of Both Model¶

Final Model¶

Predict Unseen Data¶

Conclusion¶